A method and system for evaluating driving risk of online car-hailing based on driver state monitoring

Abstract

The application discloses a kind of based on the driving risk assessment method and system of driver state monitoring of online car-hailing, it is related to driving safety monitoring and online car-hailing operation safety management technical field.The method obtains driving state data, vehicle data, operation data, environmental data and historical driving data;Determine operation task and task related state change, obtain state recovery feature, and determine state recovery deviation data in combination with personal recovery baseline;When recovery ability drops, determine target driving demand and available response time, further determine driving response time and driving response margin;When driving response margin reaches risk condition, execute driving state encryption monitoring, and generate driving risk assessment result in combination with state recovery change.The application can reduce the interference of normal operation behavior on driving state judgment, and improve the pertinence and reliability of driving risk assessment.

Description

Technical Field

[0001] This invention relates to the field of driving safety monitoring and ride-hailing operation safety management technology, and in particular to a method and system for assessing ride-hailing driving risks based on driver status monitoring. Background Technology

[0002] During continuous ride-hailing operations, drivers need to frequently perform tasks such as navigation confirmation, finding other drivers, observing the roadside, and parking. Continuous operation can easily lead to a decline in attention and recovery ability. Therefore, it is necessary to accurately assess driving risks by combining the driver's actual condition with the current driving task.

[0003] Existing technologies typically use in-vehicle cameras to collect driver status information such as eye opening and closing, blinking frequency, gaze direction, and head posture to identify fatigue or distraction, and combine this information with vehicle operating parameters, road environment information, or preset risk models to determine the driving risk level.

[0004] However, existing technologies still have shortcomings in practical applications: on the one hand, the gaze deviation and head rotation generated during normal ride-hailing operations overlap with the actual decline in attention, making it difficult to accurately distinguish between normal task behavior and abnormal state; on the other hand, the impact of the same degree of driver state decline on driving safety varies under different road environments and short-term driving tasks, and it is difficult to accurately characterize the actual risk corresponding to the current driving task based solely on the degree of state abnormality. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the present invention aims to provide a method and system for assessing ride-hailing driving risks based on driver status monitoring, thereby resolving at least one of the problems in the background technology.

[0006] The first aspect of the present invention is to provide a method for assessing the driving risk of ride-hailing vehicles based on driver status monitoring, the method comprising: acquiring driving status data, vehicle data, operational data, environmental data and historical driving data;

[0007] The operation task is determined based on the operation data and vehicle data, the driving status data is matched with the operation task to obtain task-related status changes, and the status recovery characteristics are determined based on the driving status data after the task-related status changes end.

[0008] A personal recovery baseline is established based on the historical driving data, and the state recovery characteristics are compared with the personal recovery baseline to obtain state recovery deviation data.

[0009] When the state recovery deviation data indicates a decline in recovery capability, the target driving demand and available response time are determined based on the environmental data and vehicle data.

[0010] The driving response time is determined based on the target driving requirements, state recovery deviation data, and historical driving data, and is compared with the available response time to obtain the driving response margin.

[0011] When the driving response margin reaches the risk condition, encrypted monitoring of the driving state is performed to obtain the state recovery change. The driving risk assessment result is generated by combining the driving response margin and the state recovery change.

[0012] Furthermore, based on the driving state data after the completion of the task-related state changes, state recovery characteristics are determined, including:

[0013] Extract forward gaze state and head posture change data from the driving state data after the completion of the task-related state changes to obtain recovery monitoring data;

[0014] Based on the recovery monitoring data, the forward gaze state and head posture changes are continuously judged to determine the time when the state recovery is completed.

[0015] The state recovery time is determined based on the end time of the task-related state change and the completion time of state recovery, and the state recovery characteristics are determined in conjunction with the recovery monitoring data.

[0016] Furthermore, based on the restored monitoring data, continuous determination is made of forward gaze status and head posture changes, including:

[0017] The first recovery moment when the driver's line of sight re-enters the forward road area is determined based on the recovery monitoring data;

[0018] Starting from the first recovery time, the forward gaze state and head posture changes are continuously monitored to determine the continuous recovery interval that meets the stable recovery conditions.

[0019] The time when state recovery is completed is determined based on the continuous recovery interval.

[0020] Furthermore, the continuous recovery interval that meets the stable recovery conditions is determined, including:

[0021] The forward gaze continuity status and head posture deviation status are determined based on the recovery monitoring data.

[0022] When the forward gaze is continuously maintained and the forward stability condition is met, and the head posture deviation condition is met, the corresponding time period is determined as a candidate recovery time period.

[0023] Consecutive candidate recovery periods are merged to determine the continuous recovery interval that meets the stable recovery conditions.

[0024] Further, the state recovery characteristics are compared with the individual recovery baseline to obtain state recovery deviation data, including:

[0025] Based on the operational tasks, vehicle data, and environmental data, a target recovery baseline is determined from the individual recovery baseline;

[0026] The recovery characteristics are compared with the target recovery baseline to obtain recovery difference data;

[0027] Based on the recovery difference data, the deviation of the current state recovery from the target recovery baseline is determined, and the state recovery deviation data is obtained.

[0028] Further, determining the deviation of the current recovery state from the target recovery baseline based on the recovery difference data includes:

[0029] Based on the recovery difference data, determine the differences in state recovery time and forward line of sight stability corresponding to continuous state recovery events;

[0030] The continuous deviation of consecutive state recovery events from the target recovery baseline is determined based on the state recovery time difference;

[0031] When the state recovery time of a continuous state recovery event is consistently higher than the baseline state recovery time corresponding to the target recovery baseline, and the forward line-of-sight stability is consistently lower than the baseline forward line-of-sight stability corresponding to the target recovery baseline, the state recovery deviation data is determined.

[0032] Further, determining the driving response time and comparing it with the available response time to obtain the driving response margin includes:

[0033] The time difference between the two is determined based on the available response time and the driving response time;

[0034] The driving response margin corresponding to the target driving requirement is determined based on the time difference.

[0035] Further, determining the driving response margin corresponding to the target driving demand based on the time difference includes:

[0036] Determine the corresponding response margin boundary based on the target driving requirements;

[0037] The time difference is compared with the response margin boundary to determine the positional relationship of the time difference relative to the response margin boundary;

[0038] The driving response margin corresponding to the target driving requirement is determined based on the positional relationship and time difference.

[0039] Furthermore, when the driving response margin reaches a risk condition, encrypted monitoring of the driving state is performed, including:

[0040] Determine the corresponding risk margin boundary based on the target driving requirements;

[0041] The driving response margin is compared with the risk margin boundary. When the driving response margin is not higher than the risk margin boundary, it is determined that the driving response margin has reached the risk condition.

[0042] After the aforementioned risk conditions are met, the sampling frequency of driving status data is increased, and the state recovery change is determined based on the encrypted driving status data.

[0043] A second aspect of the present invention provides a ride-hailing driving risk assessment system based on driver status monitoring, used to implement the method described in the first aspect, the system comprising:

[0044] The data acquisition module is used to acquire driving status data, vehicle data, operational data, environmental data, and historical driving data.

[0045] The status recovery module is used to determine operational tasks and task-related status changes, and to determine status recovery characteristics;

[0046] The state deviation module is used to establish an individual recovery baseline and determine state recovery deviation data based on the state recovery characteristics.

[0047] The driving demand module is used to determine the target driving demand and available response time based on the environmental data and vehicle data.

[0048] The response analysis module is used to determine the driving response time and obtain the driving response margin based on the driving response time and available response time.

[0049] A risk assessment module is used to perform encrypted monitoring of the driving state and generate a driving risk assessment result when the driving response margin reaches a risk condition. A third aspect of the invention provides a computer-readable storage medium storing one or more programs that, when executed by a processor, implement the method described in the first aspect.

[0050] A fourth aspect of the present invention provides an electronic device comprising a memory and a processor, wherein:

[0051] The memory is used to store computer programs;

[0052] The processor is used to execute a computer program stored in the memory to implement the method described in the first aspect. Compared with the prior art, the advantages of the present invention are:

[0053] 1. This invention matches driving status data with operational tasks, determines state recovery characteristics based on the recovery process after the task-related state changes end, and then compares them with the individual recovery baseline to obtain state recovery deviation data. This reduces misjudgments caused by the confusion between gaze deviation, head rotation and actual attention decline caused by normal operational tasks, thereby improving the accuracy of driver status recognition.

[0054] 2. This invention determines the target driving requirements and available response time, and determines the driving response time by combining the driver's state recovery deviation and historical driving data, further obtaining the driving response margin. After the risk conditions are met, the invention generates a driving risk assessment result by combining the state recovery changes obtained from encrypted monitoring. This allows the risk assessment to reflect the degree of matching between the driver's current response capability and the actual driving task requirements, thereby improving the pertinence and reliability of the driving risk assessment. Attached Figure Description

[0055] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:

[0056] Figure 1 This is a flowchart of the ride-hailing driving risk assessment method in Embodiment 1 of the present invention.

[0057] Figure 2 This is a structural block diagram of the ride-hailing driving risk assessment system in Embodiment 2 of the present invention. Detailed Implementation

[0058] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided so that the disclosure of the present invention will be more thorough and complete.

[0059] Example 1

[0060] According to an embodiment of the present invention, a method for assessing ride-hailing driving risks based on driver status monitoring is provided. Please refer to the following embodiment. Figure 1 , Figure 1 The flowchart of a ride-hailing driving risk assessment method based on driver status monitoring provided in Embodiment 1 of the present invention is shown, specifically including steps S1 to S6.

[0061] S1: Acquire driving status data, vehicle data, operational data, environmental data, and historical driving data.

[0062] In this embodiment, driving status data, vehicle data, operational data, and environmental data are continuously acquired during ride-hailing operations, along with historical driving data corresponding to the current driver. The above data retains the collection time or the event occurrence time, and is mapped according to a unified time benchmark, so that changes in the driver's state can be correlated with the vehicle's operating status, operational tasks, and road environment at the corresponding time.

[0063] Specifically, driving state data can be acquired by a driver monitoring camera device installed in the cockpit. By recognizing the driver's facial images of eye state, gaze direction, and head posture, the following can be obtained: eye opening and closing status, duration of eye closure, gaze direction, duration of gaze away from the forward road area, head deflection status, and forward gaze maintenance status. This driving state data is used to subsequently identify task-related state changes that occur during the driver's operational tasks and to further analyze the state recovery after the completion of these task-related state changes.

[0064] Vehicle data can be acquired from the vehicle's CAN bus, vehicle controller, and body controller, including vehicle speed, longitudinal acceleration, lateral acceleration, steering wheel angle, braking status, acceleration status, turn signal status, and vehicle position. This vehicle data is used, on the one hand, in conjunction with operational data to determine the current operational task, and on the other hand, to subsequently combine with environmental data to determine target driving requirements and available response time.

[0065] Operational data can be obtained from ride-hailing operation terminals, operation platform interfaces, and in-vehicle navigation terminals, including order status, pick-up location, passenger boarding status, passenger carrying status, destination location, and navigation prompts. The order status indicates the current operational stage of the vehicle, while the navigation prompts indicate the vehicle's planned direction and approach to a turning point. After correlating the operational data with the vehicle's location, speed, and turning status over time, it is used to subsequently determine the driver's current operational task.

[0066] Environmental data can be acquired by the vehicle's forward-facing camera, millimeter-wave radar, in-vehicle navigation map, and vehicle environmental perception device. This data includes distance to the vehicle ahead, relative speed to the vehicle ahead, pedestrian and non-motorized vehicle targets, lane status, intersection location, road curvature, road speed limits, and planned driving route. This environmental data is primarily used to determine, in conjunction with vehicle data, the driving demands requiring a response from the driver within a short timeframe ahead of the vehicle after the driver's ability to recover has diminished.

[0067] Historical driving data refers to the driving records corresponding to the current driver and can be stored in the vehicle's storage unit or the ride-hailing operation safety platform. Historical driving data includes at least historical driving states, vehicle operating states, corresponding operational tasks, road conditions, state recovery records, and driving operation response records. Among these, historical records generated under normal driving conditions are used to establish an individual recovery baseline; historical response records corresponding to driving operations such as braking and steering are used to determine driving response times by combining target driving needs and state recovery deviation data.

[0068] To ensure comparable time relationships between data from different sources, a unified vehicle clock or existing vehicle timing information can be used as a unified time reference to associate driving status data, vehicle data, operational data, and environmental data according to timestamps. For periodic data with different sampling frequencies, valid data near the same time can be selected for matching; for continuous status data such as order status and turn signal status, data that is valid at the corresponding time is used for matching.

[0069] By acquiring and matching the above data with time, it becomes possible to determine whether changes in driving status are task-related based on actual operational conditions, and to provide a data foundation for determining status recovery characteristics, individual recovery baseline, target driving needs, driving response time, and driving response margin.

[0070] S2: Determine the operation task based on the operation data and vehicle data, match the driving status data with the operation task to obtain task-related status changes, and determine the status recovery characteristics based on the driving status data after the task-related status changes end.

[0071] Specifically, the operational task is used to characterize the road observation and vehicle operation tasks that the driver needs to complete in the current operational phase of the ride-hailing service, combining order status, navigation prompts, and the actual vehicle operation. In this embodiment, the operational task is not determined solely based on order status or navigation prompts. Instead, the task requirements reflected in the operational data are mapped to the actual operating status reflected in the vehicle data to reduce the impact of navigation pre-prompts, early order status switching, and other factors on the identification of operational tasks.

[0072] The order status, pick-up location, destination location, and navigation prompts are extracted from the operational data, and combined with the vehicle location, vehicle speed, turn signal status, steering wheel angle, and vehicle motion status from the vehicle data to determine the current operational task.

[0073] For example, when the navigation system indicates that the vehicle is approaching a turning position, and the vehicle actually approaches the corresponding turning position and decelerates, while the turn signal in the corresponding direction is activated or the steering wheel turns in the corresponding direction, it is determined that there is a navigation confirmation and turning observation task. When the order status indicates that the vehicle is in the pick-up stage, and the vehicle has entered the vicinity of the pick-up location and is maintaining a low speed, it is determined that there is a passenger search and pick-up observation task. When the order status indicates that the vehicle is in the passenger-carrying stage, and the vehicle is gradually approaching the order destination and decelerating or moving towards the side of the road, it is determined that there is a destination confirmation and parking task. When the vehicle activates its turn signal and shows a lateral movement trend in the corresponding direction, it is determined that there is a lane change observation task.

[0074] For different operational tasks, corresponding task observation areas and effective time periods are determined. The task observation area represents the road direction that the driver needs to normally observe when completing the corresponding operational task. For example, the task observation area for a right turn observation task may include the right rearview mirror area and the road area to the right front of the vehicle; the left turn observation task corresponds to the left rearview mirror area and the road area to the left front of the vehicle; the pick-up observation task corresponds to the roadside area and the area in front of the vehicle where the pick-up location is located; and the parking task corresponds to the rearview mirror area, the roadside area, and the area in front of the vehicle on the side of the target parking spot. The location of the pick-up location or destination on which side of the vehicle's driving direction is located can be determined based on the spatial relationship between the vehicle's current position, the planned driving route, and the target location.

[0075] The effective time period for a task is used to define the time range within which the driver is actually required to perform the corresponding observation actions for the current operational task. The effective time period can be determined based on the spatial relationship between the vehicle and the task target location, as well as the vehicle's current operating status. For example, for a turning task, it can start from the corresponding range before the vehicle enters the turning node and continue until the vehicle completes the corresponding turning operation; for a pick-up task, it can start from the time the vehicle enters the vicinity of the pick-up location and continue until the vehicle completes the pick-up and stops or leaves the pick-up area. The effective time periods for different operational tasks can be determined based on vehicle speed, road type, and the actual duration of the corresponding task, without needing to adopt a uniform fixed duration.

[0076] Furthermore, changes in the driver's gaze are identified from the driving status data. Based on the change in the driver's gaze direction relative to the forward road area, it is determined whether the driver's gaze has shifted from the forward road area to the task observation area corresponding to the current operational task, and the direction of the gaze change can be further confirmed by combining the direction of head turn.

[0077] When the change in gaze occurs within the effective time period of the current operational task, and the direction of gaze deviation matches the observation area of ​​the current operational task, the change in driving state is determined to be a task-related state change.

[0078] For example, when a vehicle is performing a right turn, if the driver shifts their gaze from the front of the vehicle to the right rearview mirror or the road area to the right as they approach an intersection, this change in gaze corresponds to the right turn task in both time and direction of observation. Therefore, it is identified as a task-related state change, and not simply as driver distraction due to the driver's gaze temporarily leaving the forward road area. Changes in gaze occurring outside the effective time frame of the task, or in directions where the gaze shift does not correspond to the observation direction required for the current operational task, are not identified as task-related state changes for the current operational task.

[0079] After identifying a task-related state change, the termination state of that change is determined. When the driver's gaze leaves the task observation area corresponding to the current operational task and begins to move back towards the road area, the task-related state change is determined to have ended, and the moment of termination is recorded. This establishes a clear temporal boundary between the end of the task-related state change and the subsequent recovery process of the driving state.

[0080] Starting from the end of the task-related state change, driving state data continues to be acquired. Forward gaze state and head posture change data are extracted from this driving state data to obtain recovery monitoring data. This recovery monitoring data characterizes the process by which the driver recovers from a task observation state to a stable forward road attention state after completing the corresponding operational task observation.

[0081] To assess the recovery process, the forward road area and normal forward posture range can be pre-determined based on the installation location of the driver monitoring device, the cockpit structure, and the driver's normal forward observation state. The forward road area is the visual field corresponding to the driver's normal observation of the road ahead through the vehicle's windshield; the normal forward posture range characterizes the range of the driver's head posture while maintaining normal forward road observation. The normal forward posture range is primarily used to determine whether the driver's head posture has recovered after the task observation period ends, and is not used as a replacement for subsequent personal recovery baselines.

[0082] Based on the recovery monitoring data, the moment when the driver's gaze re-enters the forward road area after the task-related state change is determined is defined as the first recovery moment. It should be noted that the first recovery moment only indicates that the driver's gaze has returned to the forward road area; it does not directly indicate that the driver has re-established a stable forward attention state. Therefore, continuous assessment of the driving state after the first recovery moment is still necessary.

[0083] Starting from the first recovery time, the forward gaze continuity status and head posture deviation status are determined based on continuously acquired recovery monitoring data. The forward gaze continuity status indicates whether the driver's gaze remains continuously within the forward road area; the head posture deviation status indicates the deviation of the current head posture from the normal forward posture range.

[0084] Within a single continuous sampling period, when the driver's line of sight remains on the forward road area, it is determined that the continuous forward line of sight meets the forward stability condition; when the driver's head posture is within the normal forward posture range, it is determined that the head posture deviation meets the posture recovery condition.

[0085] When the forward gaze is continuously maintained and meets the forward stability condition, and the head posture deviation meets the posture recovery condition, the corresponding time period that simultaneously meets both conditions is determined as the candidate recovery time period.

[0086] Temporally consecutive candidate recovery periods are merged. If there is no driving state between adjacent candidate recovery periods that would cause the failure of forward stabilization or attitude recovery conditions, then adjacent candidate recovery periods are grouped into the same continuous recovery process. When the duration of the continuous recovery process reaches the observation duration required for stable recovery, the corresponding continuous recovery process is determined as a continuous recovery interval that satisfies the stable recovery conditions.

[0087] The observation duration required for stable recovery is designed to prevent the driver's line of sight from only momentarily passing over the forward road area, or their head posture from only briefly entering the normal forward posture range, from being judged as having completed state recovery. In practical applications, this continuous judgment can be achieved by ensuring that the forward stability and posture recovery conditions are met at multiple consecutive valid sampling moments, based on the sampling frequency of the driver monitoring device. The specific observation duration can be calibrated according to the sampling capability of the driver monitoring device and the actual vehicle monitoring needs.

[0088] The state recovery completion time is determined based on the continuous recovery interval. Specifically, when the continuous recovery process formed after the first recovery time reaches the observation duration required for stable recovery, the time corresponding to the reached observation duration is determined as the state recovery completion time. In this way, the driver's first re-entry into the forward road area is distinguished from the driver's re-establishment of a stable forward attention state.

[0089] Furthermore, the state recovery time is determined based on the end time of the task-related state change and the completion time of state recovery. The state recovery time is the time elapsed from the end of the task-related state change to the driver regaining a stable forward attention state, used to characterize the speed at which the driver regains forward road attention after completing normal operational task observation. An increase in the state recovery time indicates that the time required for the driver to re-establish stable forward attention after completing the corresponding task observation has increased.

[0090] Simultaneously, the stability of forward visual line of sight is determined based on the recovered monitoring data. Specifically, this can be determined by assessing whether the driver's gaze remains within the forward road area from the first recovery moment to the moment the recovery is complete. For example, the proportion of valid sampled data where the gaze is within the forward road area during this time frame can be statistically analyzed, and the stability of forward visual line of sight is characterized by considering the frequency with which the gaze leaves the forward road area during the recovery process. The more stably the driver's gaze remains within the forward road area during the recovery process, the higher the stability of forward visual line of sight.

[0091] A state recovery feature is formed based on the state recovery time and the degree of forward gaze stability. Head posture change data is primarily used, along with forward gaze status, to determine the moment state recovery is complete. State recovery time and the degree of forward gaze stability serve as the main indicators in this embodiment for subsequently determining whether the driver's state recovery ability deviates from their normal state.

[0092] For example, when a driver performs a right turn observation task, they shift their gaze from the forward road area to the right-side rearview mirror area. After completing the right-side observation, they begin to shift their gaze back to the front of the vehicle. The end time of the task-related state change is recorded at this point. The first recovery time is determined when the driver's gaze first re-enters the forward road area. However, if the driver subsequently deviates briefly from the forward road area, the state recovery is not immediately determined. Only when the driver's forward gaze state and head posture continuously meet the corresponding recovery conditions and form a continuous recovery interval that meets the observation duration requirement is the state recovery completion time determined. Based on this, the state recovery time and the stability of the forward gaze corresponding to the task-related state change are obtained.

[0093] Through the above processing, reasonable changes in line of sight caused by operational tasks such as navigation confirmation, turning observation, finding the driver, and parking are not directly judged as abnormal driver states. Instead, the operational tasks are first used to correlate these state changes. After confirming that they are task-related state changes, the recovery process after the task-related state change ends is further analyzed. This shifts the focus of subsequent state judgment from single line-of-sight deviations to the driver's ability to re-establish stable forward attention after completing normal operational tasks. This provides a data foundation for the next step of comparing these state recovery characteristics with the individual recovery baseline and determining state recovery deviation data.

[0094] S3: Establish a personal recovery baseline based on the historical driving data, and compare the state recovery characteristics with the personal recovery baseline to obtain state recovery deviation data.

[0095] Specifically, the personal recovery baseline is used to characterize the level of recovery a driver typically exhibits when re-establishing stable forward attention after completing different operational tasks under normal driving conditions. Considering individual differences among drivers in rearview mirror observation habits, gaze shift speed, and head movement amplitude, and also the variations in the normal recovery process of the same driver under different operational tasks, vehicle operating states, and road conditions, this embodiment does not use a single fixed recovery threshold shared by all drivers. Instead, it filters normal driving records from the historical driving data corresponding to the current driver and establishes a personal recovery baseline based on the corresponding driving conditions.

[0096] First, normal state recovery records for establishing an individual recovery baseline are selected from the historical driving data. The historical driving data includes at least historical driving states, vehicle operating states, corresponding operational tasks, road conditions, and state recovery records corresponding to changes in state related to each task. For each historical task-related state change, the corresponding historical state recovery time and historical forward line-of-sight stability are determined using the same state recovery determination method as in step S2.

[0097] To avoid abnormal driving processes affecting an individual's recovery baseline, historical recovery records can be filtered for normal states. Specifically, if there are no obvious abrupt changes in vehicle operation status within the corresponding historical period, the driver does not exhibit continuous abnormal eye closure, and the corresponding operational task can be completed normally, the historical recovery record will be considered a normal recovery record. For example, for a normal steering observation task, the corresponding vehicle can be required to complete a normal deceleration and steering process without any emergency braking or obvious abnormal steering caused by the driver's delayed response; for pick-up or parking tasks, the vehicle can be required to complete the deceleration and parking normally according to the corresponding operational task. Historical records that are clearly accompanied by abnormal driving operations or abnormal driving states will not be used to establish an individual's recovery baseline.

[0098] After obtaining the normal state recovery records, they are aggregated according to the corresponding operational tasks, vehicle operating status, and road environment. In specific implementation, the normal state recovery records can first be classified according to the type of operational task, such as forming historical recovery data corresponding to different tasks like turning observation, lane change observation, pick-up observation, and parking; then, historical recovery data corresponding to the same type of operational task can be further aggregated according to the vehicle speed range; for situations where differences in road environment have a significant impact on the driver's normal observation behavior, environmental data such as road type, intersection status, and roadside target conditions can also be combined for differentiation.

[0099] For example, during a low-speed pick-up operation, the same driver needs to continuously observe the side of the road and the passenger's position, and their recovery time may differ from the recovery time after a turn on a normal urban road. Similarly, even with the same right-turn observation task, the normal recovery process of a driver shifting from lateral observation back to stable forward attention may differ under low-speed side road turning conditions versus higher-speed road turning conditions. Therefore, by aggregating historical normal recovery data according to operational tasks, vehicle operating status, and environmental conditions, it is possible to avoid directly using recovery data generated under significantly different driving conditions as a unified comparison benchmark.

[0100] For each type of normal state recovery record after aggregation, a baseline state recovery time and a baseline forward gaze stability are determined. The baseline state recovery time characterizes the time typically required for a driver to recover to stable forward attention after completing a task-related state change under corresponding driving conditions; the baseline forward gaze stability characterizes the normal level of stability of the driver's forward gaze during the recovery process under the same conditions.

[0101] In practice, the recovery time of similar normal recovery records can be statistically analyzed, and the concentrated interval or statistical representative value that can represent the driver's normal recovery level can be used as the baseline recovery time. Correspondingly, the stability of forward vision can be statistically analyzed to determine its normal range or statistical representative value as the baseline forward vision stability. For cases with a large number of historical records, the median, quantile interval, or the mean after removing obvious outliers can be used to represent the normal level, reducing the impact of occasional changes in state on the individual's recovery baseline. The above statistical methods represent a stable representation of normal historical driving records, aiming to obtain a reference for the driver's normal recovery under the same or similar driving conditions, without changing the technical logic of this application, which is based on comparing individual historical recovery states.

[0102] The baseline recovery time and baseline forward line-of-sight stability corresponding to different operational tasks, vehicle operating states, and road environmental conditions are correlated and stored to form the personal recovery baseline. Therefore, the personal recovery baseline is not a single value, but rather a set of reference values ​​for normal state recovery formed by the current driver under different driving conditions.

[0103] After obtaining the state recovery characteristics determined in step S2 during the current driving process, a target recovery baseline corresponding to the current driving conditions is determined from the personal recovery baseline based on the current operational tasks, vehicle data, and environmental data.

[0104] Specifically, firstly, the corresponding task category is determined based on the current operational task, and normal recovery records of the same task category are filtered in the personal recovery baseline; then, based on the current vehicle speed, vehicle operating status and road environment, the baseline record that is the same as or similar to the current driving conditions is determined in the filtered recovery records, and the corresponding baseline state recovery time and baseline forward line of sight stability are determined as the target recovery baseline.

[0105] For example, if the current state recovery characteristics arise from a right-turn observation task on an urban road, and the vehicle is within the corresponding speed range and the current road is a typical urban intersection, then the target recovery baseline is preferentially selected from the individual recovery baseline based on the normal recovery data generated by the driver under similar right-turn tasks, speed conditions, and road conditions. If no identical historical conditions exist, the target recovery baseline can be determined from historical recovery records with minimal differences in vehicle speed and road conditions, while maintaining consistency in the operational task type. This allows for comparison of the current state recovery characteristics with comparable individual historical normal recovery states.

[0106] Furthermore, the state recovery characteristics are compared with the target recovery baseline to obtain recovery difference data.

[0107] Specifically, the state recovery time in the current state recovery feature is compared with the baseline state recovery time corresponding to the target recovery baseline to obtain the state recovery time difference; the current forward line of sight stability is compared with the baseline forward line of sight stability corresponding to the target recovery baseline to obtain the forward line of sight stability difference.

[0108] For example, when the baseline recovery time corresponding to the target recovery baseline is within a normal time range, the difference in recovery time can be determined based on the degree of deviation of the current recovery time from the upper limit of that normal time range. When the current recovery time is within the normal range, the time difference indicates that there is no significant extension. When the current recovery time exceeds the normal range, the difference in recovery time is characterized by the degree to which it exceeds the normal range. For forward line-of-sight stability, the difference can be formed according to the degree of reduction in the current forward line-of-sight stability relative to the baseline forward line-of-sight stability or the normal stability range.

[0109] It should be noted that an increase in the time required for a single state recovery does not necessarily indicate a real decline in the driver's state recovery ability. For example, a driver may experience a short-term recovery delay due to complex road information. Therefore, this embodiment further utilizes continuous state recovery events to continuously assess the recovery difference data, thereby reducing misjudgments caused by single, accidental fluctuations.

[0110] Specifically, during continuous operation, the differences in state recovery time and forward visibility stability corresponding to each state recovery event are retained according to the chronological order of their occurrence. The continuous state recovery events refer to multiple task-related state change events that occur sequentially during continuous operation and whose state recovery characteristics can be determined according to step S2. These events can originate from consecutively occurring steering observations, rearview mirror observations, pick-up observations, or other normal operational observation tasks, and do not require that each state recovery event belong to the exact same operational task. For different operational tasks, a target recovery baseline corresponding to their respective driving conditions is used for comparison, ensuring that the resulting recovery differences are comparable relative to their respective normal states.

[0111] Based on the differences in state recovery time, it is determined whether consecutive state recovery events consistently exhibit recovery times exceeding the corresponding target recovery baseline. When the state recovery time corresponding to multiple consecutive state recovery events is consistently higher than the baseline state recovery time in their respective target recovery baselines, it is determined that a continuous deviation in state recovery time has occurred.

[0112] Simultaneously, the driver's forward visual stability is determined based on the differences in forward visual stability. When the forward visual stability corresponding to a continuous recovery event is consistently lower than the baseline forward visual stability in its respective target recovery baseline, it indicates that the driver's stability in re-establishing stable forward attention after the task observation has decreased.

[0113] When the state recovery time of a continuous state recovery event is consistently higher than the baseline state recovery time corresponding to the target recovery baseline, and the forward gaze stability is consistently lower than the baseline forward gaze stability corresponding to the target recovery baseline, it is determined that the current state recovery has formed an effective deviation from the individual's normal recovery state, and state recovery deviation data is obtained accordingly.

[0114] The state recovery deviation data includes at least the state recovery time, state recovery time difference, forward gaze stability, and forward gaze stability difference corresponding to consecutive state recovery events, and records the persistence of the above deviations in consecutive state recovery events. Therefore, the state recovery deviation data is not based on a single gaze deviation or a single eye-closing event, but rather on state recovery characteristics that have excluded the influence of reasonable observation during operational tasks, and is further confirmed through the driver's own normal recovery baseline and consecutive state recovery events.

[0115] For example, suppose a driver's recovery time under corresponding steering and rearview mirror observation conditions is mainly distributed between 0.6 and 0.8 seconds, and the forward gaze is highly stable during the normal recovery process. If the recovery times for the current three consecutive recovery events are 1.4 seconds, 1.5 seconds, and 1.6 seconds, respectively, and the stability of the forward gaze is lower than the respective target recovery baseline during all three recovery processes, then it can be determined that the current driver's recovery process deviates continuously from the individual's normal recovery level. The above values ​​are only used to illustrate the process of determining the recovery deviation data and do not constitute a limitation on the actual judgment threshold or applicable drivers.

[0116] Through the above processing, this step establishes a personal recovery baseline using the driver's historical normal driving data, and determines a comparable target recovery baseline based on the current operational task, vehicle status, and environmental conditions. Then, it uses the difference between the current state recovery characteristics and the target recovery baseline, as well as the continuous deviation of consecutive state recovery events, to form state recovery deviation data. Therefore, the state recovery deviation data can characterize the change in the driver's actual state recovery ability relative to their normal level after completing a normal operational task. This data is used in the next step, when the state recovery deviation data indicates a decline in recovery ability, to further determine the target driving needs and available response time by combining road environment and vehicle data.

[0117] S4: When the state recovery deviation data indicates a decline in recovery capability, the target driving demand and available response time are determined based on the environmental data and vehicle data.

[0118] Specifically, this step is executed after step S3 has confirmed that the driver's recovery ability has decreased relative to their normal recovery level. The state recovery deviation data indicates a decrease in recovery ability, meaning that the recovery time for consecutive state recovery events is consistently higher than the baseline state recovery time in the corresponding target recovery baseline, while the forward visual stability is consistently lower than the baseline forward visual stability in the corresponding target recovery baseline. At this point, it only indicates that the driver's ability to re-establish stable forward attention has decreased; it does not directly indicate that the same level of driving risk has been established. Therefore, it is necessary to determine the driver's response capability requirements based on the actual road driving task currently being faced by the vehicle.

[0119] In this embodiment, the short-term driving demand observation range ahead of the vehicle is determined based on the vehicle position, vehicle speed, and vehicle motion state in the vehicle data, as well as the forward target, road structure, lane status, and planned driving path in the environmental data. This short-term driving demand observation range is used to define the spatial or temporal range within which driving demand analysis needs to be performed after the current state's recovery capability declines. This ensures that subsequent judgments are based on the actual driving task the vehicle is about to face, rather than performing an indiscriminate analysis of the distant road environment.

[0120] In one implementation, the expected travel range within a certain future timeframe can be determined based on the vehicle's current speed, and this expected travel range can be used as the driving demand observation range. For example, the road range that the vehicle can reach within the next 5 to 15 seconds based on its current driving state can be used as an exemplary driving demand observation range. For higher vehicle speeds, the corresponding spatial observation distance increases accordingly; for lower speeds such as picking up other vehicles or passing through intersections at low speeds, the corresponding spatial observation distance decreases accordingly. The aforementioned 5 to 15 seconds is only an exemplary range for illustrating this implementation method; in actual applications, it can be set according to road type, vehicle speed, and the vehicle's environmental perception range.

[0121] Within the scope of the driving demand observation, candidate driving demands requiring the driver to perform road observation or vehicle control are identified based on environmental and vehicle data. These candidate driving demands may include deceleration or braking demands corresponding to the deceleration of the vehicle in front, braking or avoidance demands corresponding to pedestrians or non-motorized vehicles entering the vehicle's expected driving area, observation and traffic control demands corresponding to the vehicle approaching an intersection, deceleration and steering demands corresponding to the vehicle approaching a curved road section, and steering or lane adjustment demands required according to the planned driving path.

[0122] It should be noted that the candidate driving requirements in this step are not generated solely based on the presence of a target in the environment, but rather determined by the relationship between the environmental target and the vehicle's current state of motion. For example, even if there are other vehicles ahead, a stable distance with no tendency to approach does not necessarily create a candidate driving requirement for the driver to brake; however, if the distance between the vehicle and the vehicle in front continuously decreases, or if the vehicle in front has already decelerated while the vehicle maintains its original driving state, then a candidate driving requirement corresponding to the deceleration of the vehicle in front may be created. Similarly, for pedestrians or non-motorized vehicles, it is necessary to determine whether they are likely to enter the vehicle's driving conflict zone based on the target's location, the direction of movement reflected in continuous environmental data, and the vehicle's expected driving trajectory, before determining whether a corresponding avoidance requirement has been created.

[0123] For each candidate driving requirement, the corresponding driving operation that the driver needs to perform is determined. This driving operation is a component of the target driving requirement itself, primarily representing the observation or vehicle control behaviors the driver needs to implement in response to corresponding road conditions. For example, a vehicle decelerating in front corresponds to a deceleration or braking operation; a pedestrian entering the vehicle's expected driving area corresponds to a braking or avoidance operation; a curved road section corresponds to a deceleration and steering operation; and a lane adjustment requirement corresponds to lateral road observation and steering operations. This correspondence is used in subsequent step S5 to extract historical response records of the same or similar driving operations from historical driving data based on the target driving requirement.

[0124] Furthermore, for each candidate driving requirement, the corresponding available response time is determined based on the distance of the environmental target or road position relative to the vehicle, the vehicle's current speed, and the relative motion relationship between the vehicle and the relevant target. The available response time refers to the time range from the current analysis moment until the current road conditions require the driver to begin implementing the corresponding driving response; it characterizes the objective time constraint of the current driving task on the driver's response speed.

[0125] Specifically, for different types of candidate driving needs, the response constraint position or response constraint state can be determined based on their corresponding road constraint relationships.

[0126] For driving demands arising from the deceleration of the vehicle ahead, the time required for the distance between the current vehicle and the vehicle ahead to reach the corresponding safe following constraint can be predicted based on the distance between them, the relative speed of the vehicle ahead relative to the current vehicle, and the distance change trend reflected in continuous environmental data. This prediction allows the current driver to determine the available response time to initiate deceleration or braking. When there is no trend of the current vehicle approaching the vehicle ahead, the simply existing target vehicle ahead is not considered a driving demand with an urgent time constraint.

[0127] For driving demands arising from a vehicle approaching an intersection, curve, or turning position within a planned driving path, the estimated time required for the vehicle to reach the corresponding road position can be determined based on the distance along the planned driving path from the vehicle's current position to the corresponding road control position, combined with the vehicle's current speed and its changes. This estimated time is used to determine the available response time for the driver to complete observation, deceleration, or initiation of steering before entering the corresponding road position.

[0128] For candidate driving needs arising from pedestrians or non-motorized vehicles, the positional change of the target relative to the vehicle's expected driving area can be determined based on continuous environmental data. This, combined with the vehicle's current position, speed, and target movement trend, can be used to determine if a trajectory conflict exists. When a conflict trend exists, the available response time for the driver to brake or evade is determined by the time relationship between the vehicle and the target approaching the corresponding conflict area. When the target's movement direction continuously moves away from the vehicle's expected driving area, it is not considered a current target driving need.

[0129] Therefore, each candidate driving request has an available response time corresponding to the current road conditions. This available response time is not a fixed driving reaction threshold, but rather changes dynamically with the distance between the vehicle and road targets, vehicle speed, and relative motion. For example, under otherwise similar conditions, the closer the vehicle is to a decelerating vehicle ahead and the greater the relative approach speed, the shorter the available response time; conversely, the farther the vehicle is from an intersection or turning point, the more time the driver typically has to complete the corresponding driving response.

[0130] When there is only one candidate driving demand that requires a driver's response within the observation range of driving demand, the candidate driving demand is identified as the target driving demand, and its corresponding available response time is identified as the available response time corresponding to the target driving demand.

[0131] When multiple candidate driving needs exist simultaneously within the observation range, the available response times corresponding to each candidate driving need are compared. The candidate driving need with the shorter available response time and the more direct time constraint on the driver's current response is prioritized as the target driving need. If the available response times of multiple candidate driving needs are similar, the determination can be further based on the driving operations required for each candidate driving need. Candidate driving needs requiring explicit vehicle control operations such as braking, evasive maneuvers, and steering are prioritized, thus ensuring that the determined target driving need reflects the actual driving task with strong constraints on the driver's response ability within the current short-term range.

[0132] For example, in an exemplary operating scenario, the vehicle's forward-looking environmental perception device detects that a vehicle ahead is beginning to decelerate while the vehicle is approaching a road area where a pedestrian may cross. The system determines the available response time for two candidate driving requests based on the distance to the vehicle ahead, the relative speed of the vehicle ahead, and the positional relationship between the pedestrian and the vehicle's expected travel area. If the braking request corresponding to the deceleration of the vehicle ahead has a shorter available response time, then the braking request corresponding to the deceleration of the vehicle ahead is determined as the current target driving request. In this embodiment, the available response time corresponding to the target driving request is approximately 2.6 seconds. This value is only used to illustrate the process of determining the target driving request and the available response time and does not constitute a limitation on actual vehicle operating parameters or judgment thresholds.

[0133] Finally, this step outputs the target driving requirement and its corresponding available response time. The target driving requirement at least characterizes the actual driving task the driver needs to complete within a short timeframe and the corresponding driving operation type, which is used in the next step to determine the driver's maneuvering response based on historical driving data. The available response time characterizes the objective time conditions that allow the driver to respond to the current road driving task and is compared with the driver's driving response time in the next step to further determine the driving response margin.

[0134] Through the above processing, when it has been confirmed that the driver's ability to recover has declined, the driving risk is not determined directly based on the degree of deviation in recovery. Instead, the target driving requirements and available response time are determined based on the actual road environment and vehicle movement status. This establishes a basis for subsequent comparison between the driver's own state changes and the time requirements corresponding to the current driving task.

[0135] S5: Determine the driving response time based on the target driving requirements, state recovery deviation data, and historical driving data, and compare it with the available response time to obtain the driving response margin.

[0136] Specifically, the driving response time characterizes the time required for the driver to re-establish effective forward road attention and further formulate effective vehicle control operations corresponding to the target driving needs, starting from the current analysis time, given the driver's current recovery capability. The available response time is the length of time determined in step S4 based on the current vehicle operating state and road environment, from the current analysis time to the latest time when the corresponding driving operation needs to be effectively implemented. Therefore, the driving response time and available response time use the same starting time and time scale, allowing for direct comparison.

[0137] First, the driver's current attention recovery time is determined based on the state recovery deviation data. The attention recovery time characterizes the estimated time required for the driver to re-establish effective forward road attention under the current state recovery capability. The state recovery deviation data generated in step S3 includes the deviation of the state recovery time corresponding to consecutive state recovery events from the individual's normal recovery level. Therefore, the current attention recovery time can be determined based on the state recovery time of the most recent consecutive state recovery events and their continued deviation.

[0138] In one implementation, when the recovery times of multiple recent consecutive recovery events are all higher than the corresponding baseline recovery times, the recovery time that best represents the driver's current recovery level can be selected from the consecutive recovery events as the current attention recovery time. For example, the median value of the recovery times corresponding to the most recent recovery events can be used to reduce the impact of occasional fluctuations in a single recovery event. When consecutive recovery events exhibit a relatively clear and sustained prolonged state, the representative role of more recent recovery events in determining the current attention recovery time can also be improved, so that the obtained attention recovery time can reflect the driver's current recovery ability.

[0139] It should be noted that the current attention recovery time is used to characterize the driver's ability to recover attention during the current continuous operation, and does not mean that the driver's gaze is necessarily off the road at the current analysis moment. This time, as part of the driver's current response capability, is used to estimate the time required for the driver to regain effective road attention when a target driving demand requiring an actual response occurs.

[0140] Furthermore, the corresponding driving operation type is determined based on the target driving requirement. The target driving requirement determined in step S4 is used to characterize the actual driving tasks that the driver needs to complete first within the short-term range of the current vehicle. Therefore, the main vehicle control operations to be implemented can be determined according to the road conditions corresponding to the target driving requirement. For example, when the current vehicle is continuously approaching a decelerating vehicle ahead, the corresponding target driving requirement may be a deceleration or braking requirement; when pedestrians or non-motorized vehicles tend to enter the vehicle's expected driving area, the corresponding target driving requirement may be a braking or avoidance requirement; when the vehicle is about to enter the planned turning position, the corresponding target driving requirement may be an observation, deceleration, and turning requirement.

[0141] Based on the driving operation type corresponding to the target driving requirement, historical driving response records with the same or similar driving operation type are selected from the current driver's historical driving data. Historical driving response records used to determine the control response time are preferentially selected from those where the driver was in a normal driving state, the corresponding driving task was completed normally, and no abnormal emergency operations occurred, in order to reduce the impact of historical fatigue or abnormal control behavior on the driver's normal control response level.

[0142] For braking-related driving needs, the time between when the driver has developed effective forward road awareness and when effective braking is initiated can be extracted from historical driving data. The effective braking action can be confirmed based on brake pedal status, brake pressure changes, or vehicle data that characterizes the vehicle's actual entry into deceleration control.

[0143] For steering-related driving needs, the time between the driver completing their observation of the corresponding road and the steering wheel producing an effective turning angle consistent with the target steering direction can be extracted. For obstacle avoidance-related driving needs, the corresponding maneuver response records are extracted based on the actual braking or steering avoidance methods used during historical driving.

[0144] When multiple matching driving response records exist in historical driving data, they can be further matched according to driving operation type, vehicle speed range, and road conditions. A statistically representative value that characterizes the current driver's normal operating level can be used to determine the control response time. For example, the median value after removing obviously abnormal records can be used. If the number of historical records completely consistent with the current driving conditions is insufficient, historical records with similar vehicle operating conditions can be selected while maintaining consistency in driving operation type. However, response records of different operation types such as braking and steering should not be directly mixed to determine the same control response time.

[0145] This yields the attention recovery time corresponding to the current driver state and the maneuver response time corresponding to the target driving requirement. The driving response time is then determined based on these attention recovery and maneuver response times. The driving response time includes at least the time required for the driver to re-establish effective forward road attention and the time required to implement corresponding vehicle control after effective road attention is established.

[0146] In one implementation, the driving response time can be determined according to the following relationship:

[0147] Driving response time = Current attention recovery time + Control response time.

[0148] The current attention recovery time is determined based on state recovery deviation data and reflects the driver's current attention recovery ability; the operation response time is determined based on the target driving needs and historical driving data and reflects the normal response speed of the driver to complete the corresponding vehicle control operation. By combining the two, the driving response time is both affected by changes in the driver's current state and corresponds to the actual driving operation that needs to be completed.

[0149] For example, in one exemplary scenario, the current target driving need is the braking need corresponding to the deceleration of the vehicle in front. Based on the driver's most recent consecutive state recovery events, their current attention recovery time is determined to be approximately 1.5 seconds; further, based on the driver's historical normal braking records, it is determined that it typically takes about 0.7 seconds from regaining effective forward road attention to initiating an effective braking operation, therefore the current driving response time is approximately 2.2 seconds. The above values ​​are only used to illustrate the process of determining the driving response time and are not fixed parameters.

[0150] After obtaining the driving response time, it is compared with the available response time determined in step S4. The available response time is the length of time from the current analysis moment to the moment when the current road conditions require the driver to begin forming a corresponding effective driving operation at the latest. It is determined by the distance between the current vehicle and the road target or road position, the vehicle speed, and the relative motion relationship. Therefore, the available response time reflects the time conditions that the current driving task side can provide, while the driving response time reflects the time required for the driver to complete the current response.

[0151] The time difference between the available response time and the driving response time is determined based on the available response time and the driving response time. In one embodiment, it can be determined according to the following relationship:

[0152] Time difference = Available response time - Driving response time.

[0153] When the time difference is positive, it indicates that there is still some time remaining after the driver completes the corresponding driving response according to the current response capability. When the time difference decreases, it indicates that the remaining time after the driver completes the current driving response is gradually decreasing. When the time difference is close to zero, it indicates that the time required for the driver to complete the current driving response is close to the available response time that the current driving task can provide. When the time difference is negative, it indicates that the time required to form an effective driving operation according to the current driver's response capability has exceeded the time that the current driving task can provide.

[0154] It should be noted that the time difference is used to directly characterize the remaining time after the driver completes the response. However, the urgency of driving operations and the normal safety margin requirements corresponding to different target driving needs are not the same. Therefore, this embodiment does not simply determine the same time difference as the same driving response margin, but further combines the response margin boundary corresponding to the target driving needs to characterize the time difference.

[0155] The corresponding response margin boundary is determined based on the target driving requirement. The response margin boundary characterizes the time margin that a driver should retain after normally completing a driving response under the corresponding target driving requirement. Different types of target driving requirements can correspond to different response margin boundaries. For example, for driving requirements that require the driver to quickly implement vehicle control, such as decelerating the vehicle in front or avoiding pedestrians, a response margin boundary adapted to this type of driving requirement can be set; for driving requirements with relatively mild time constraints, such as general steering and observation, a response margin boundary corresponding to this type of driving task is adopted.

[0156] In one implementation, the response margin boundary can be pre-calibrated based on the current driver's or similar normal driving records. Specifically, the available response time and actual driving response time corresponding to a driver completing the same type of driving task under normal conditions can be extracted to determine the remaining response time corresponding to each normal driving record. The corresponding response margin boundary is then determined based on the normal remaining response time distribution that enables the stable completion of this type of driving task. For cases where personal historical records are insufficient, a basic boundary obtained through calibration using similar vehicles and road scenarios can be used and updated after accumulating the current driver's normal historical driving data.

[0157] The time difference is compared with the response margin boundary corresponding to the target driving requirement to determine the positional relationship of the time difference relative to the response margin boundary. The positional relationship can at least characterize whether the time difference is higher than the response margin boundary, close to the response margin boundary, or lower than the response margin boundary.

[0158] The driving response margin corresponding to the target driving requirement is determined based on the positional relationship and time difference. In this embodiment, the time difference is used as the numerical value of the driving response margin, and the positional relationship of the driving response margin relative to the normal margin requirement is determined based on the comparison result between the driving response margin and the response margin boundary corresponding to the target driving requirement. Subsequent comparisons with the risk margin boundary are made using the numerical value of the driving response margin.

[0159] By employing the above method, we can avoid directly assuming the same remaining response time as the same response state under different driving needs. For example, the same 0.4-second remaining time may correspond to different actual margin levels for driving needs requiring immediate braking and avoidance versus ordinary road observation needs with more lenient time requirements. Therefore, it is necessary to determine the response margin boundary corresponding to each target driving need.

[0160] Taking the aforementioned example, if step S4 determines that the available response time corresponding to the target driving demand is approximately 2.6 seconds, while this step determines that the current driver's driving response time is approximately 2.2 seconds, then the time difference between the two is approximately 0.4 seconds. Further, a corresponding response margin boundary is determined based on this target driving demand, and the positional relationship of the 0.4-second time difference relative to the response margin boundary is compared to determine the driving response margin corresponding to the current target driving demand. The aforementioned 0.4 seconds is merely an exemplary remaining response time; whether it is in a sufficient, critical, or low margin state is determined by its relationship with the response margin boundary corresponding to the current target driving demand.

[0161] It should be further clarified that the response margin boundary in this step is used to determine the relative position and remaining degree of driving response margin, and does not directly determine whether to initiate encrypted driving status monitoring in this step. Subsequent steps will further determine the risk margin boundary used to trigger risk confirmation based on the target driving needs, and decide whether to execute encrypted status monitoring based on whether the driving response margin meets the corresponding risk conditions.

[0162] Ultimately, this step yields the driving response margin corresponding to the current target driving requirements. This driving response margin establishes a correspondence between the driver's current response capability and the available time for the current driving task, enabling subsequent driving risk assessments to reflect not only the degree of driver abnormality but also the remaining response space the driver possesses after completing the actual driving task.

[0163] S6: When the driving response margin reaches the risk condition, perform encrypted monitoring of the driving state to obtain the state recovery change, and generate a driving risk assessment result by combining the driving response margin and the state recovery change.

[0164] Specifically, the driving response margin obtained in step S5 is used to characterize the remaining response space that the driver has after fulfilling the target driving requirement under the current recovery capability. The driving response margin includes at least the remaining time between the available response time and the driving response time, and associates the positional relationship of this remaining time with the response margin boundary corresponding to the current target driving requirement. This step further determines whether the driving response margin has decreased to a level of risk requiring continuous confirmation of the driver's state. If the corresponding risk condition is reached, the intensity of driving state monitoring is increased, and a driving risk assessment result is formed based on the subsequent recovery of the driving state.

[0165] First, determine the corresponding risk margin boundary based on the current target driving requirements. This risk margin boundary characterizes the minimum remaining response time allowed for the driver to complete the corresponding driving response for the current target driving requirements. When the driving response margin drops to this risk margin boundary, it indicates that there is insufficient time buffer between the driver's current response capability and the time requirements of the current driving task, and it is necessary to further confirm whether the driver's recovery capability continues to decline.

[0166] Different driving needs can have different risk margin boundaries. For example, driving needs that require the driver to quickly brake or evade, such as a vehicle rapidly decelerating ahead or a pedestrian entering the vehicle's expected driving area, should have risk margin boundaries that are commensurate with the urgency of such driving operations. Driving needs with different time requirements, such as turning on ordinary roads or lane adjustments, should have corresponding risk margin boundaries determined based on the specific driving operation. This avoids using the same fixed risk trigger conditions for all driving needs.

[0167] In one implementation, the risk margin boundary can be pre-calibrated according to the target driving demand type. Specifically, the remaining time between the available response time and the actual driving response time can be extracted from the historical records of completing the same type of driving demand under normal driving conditions. The risk margin boundary is then determined based on the lower remaining response time maintained when the driver can stably complete the corresponding driving task. If the current driver's personal historical records are insufficient for calibration, a basic risk margin boundary obtained through pre-testing of the same type of vehicle and the same driving demand can be used, and the boundary can be updated after accumulating the current driver's normal historical driving data.

[0168] It should be noted that the response margin boundary in step S5 serves a different purpose than the risk margin boundary in this step. The response margin boundary is mainly used to determine the position of the driving response margin relative to the requirements of a normal driving task, while the risk margin boundary is used to determine whether the remaining response space has decreased to the point where encrypted status monitoring needs to be activated. Under the same target driving requirement, the risk margin boundary can be set no higher than the corresponding response margin boundary, thereby forming a continuous judgment relationship from gradually tightening the response margin to risk triggering.

[0169] Furthermore, the driving response margin is compared with the risk margin boundary corresponding to the current target driving requirement. When the driving response margin is higher than the risk margin boundary, it indicates that the driver's remaining response space has not yet reached the trigger condition for encrypted monitoring, and driving status data can continue to be acquired in the normal monitoring manner; when the driving response margin is not higher than the risk margin boundary, it is determined that the driving response margin has reached the risk condition, and encrypted monitoring of driving status is initiated.

[0170] When driving response margin is numerically represented by remaining response time, it can be directly compared with the risk margin boundary. For example, if the available response time for the current target driving need is 2.6 seconds and the current driving response time is 2.2 seconds, then the corresponding remaining response time is 0.4 seconds. If the risk margin boundary corresponding to this target driving need is higher than or equal to 0.4 seconds, then it can be determined that the current driving response margin has reached the risk condition. The above values ​​are only used to illustrate the risk condition judgment process and do not constitute a fixed limitation on specific vehicles or driving needs.

[0171] Once the risk conditions are determined to be met, the sampling frequency of driving status data is increased to continuously acquire encrypted driving status data. This encrypted monitoring refers to increasing the temporal sampling density of driver status data based on the original driving status monitoring, in order to obtain more continuous information on eye status, gaze direction, and head posture changes; it does not change the monitoring object of the driving status data itself.

[0172] For example, for a driver monitoring device with multiple image acquisition levels, a conventional sampling level can be used under normal conditions, switching to a higher sampling level once the driving response margin reaches a risk condition. For instance, for a driver monitoring device supporting 30 frames per second and 60 frames per second acquisition, the speed can be switched from 30 frames per second to 60 frames per second to improve the temporal resolution of short-term line-of-sight changes and recovery processes. The above sampling frequencies are only used to illustrate the specific implementation of encrypted monitoring; the actual sampling frequency is determined based on the hardware capabilities of the driver monitoring device and the vehicle's processing capabilities.

[0173] During encrypted monitoring, the driver's subsequent state is analyzed according to the same driving state recovery determination logic as in step S2. If the driver is currently in a state recovery process, the driver's forward gaze state and head posture changes are monitored based on the encrypted driving state data. If a new task-related state change occurs, the corresponding subsequent state recovery process is determined based on the encrypted driving state data after the task-related state change ends.

[0174] For the subsequent recovery process, the corresponding recovery time and forward visual stability are determined and compared with the baseline recovery time and baseline forward visual stability in the personal recovery baseline determined in step S3. Simultaneously, the subsequent recovery characteristics continuously obtained during the encrypted monitoring period can be continuously compared over time with the most recent recovery characteristics before reaching the risk condition to determine whether the driver's recovery ability is recovering to a normal level, maintaining the current deviation, or continuing to decline.

[0175] When the state recovery time of subsequent state recovery events during the encrypted monitoring period gradually shortens and approaches the corresponding baseline state recovery time, while the forward line of sight stability gradually increases and approaches the corresponding baseline forward line of sight stability, the state recovery change is determined to be a recovery trend.

[0176] When the state recovery time corresponding to a subsequent state recovery event continues to be higher than the baseline state recovery time, and the forward line of sight stability continues to be lower than the baseline forward line of sight stability, but does not deteriorate further relative to when the risk condition is reached, it can be determined that the state recovery change is a continuous deviation state.

[0177] When the state recovery time of subsequent state recovery events further increases and the stability of forward line of sight further decreases, it is determined that the state recovery change exhibits a continuous downward trend.

[0178] Therefore, the state recovery change is used to characterize the subsequent changes in the driver's state recovery ability as the driving response margin has reached the risk condition, rather than simply repeatedly judging whether fatigue or distraction exists at a certain moment.

[0179] During the encrypted monitoring process, the subsequent state recovery characteristics are redefined based on the encrypted driving state data, and the state recovery deviation data is updated in conjunction with the personal recovery baseline. At the same time, the target driving demand, available response time, driving response time and driving response margin are redefined in accordance with steps S4 and S5, based on the latest environmental data, vehicle data and historical driving data.

[0180] The driving risk assessment results are generated by combining the driving response margin and state recovery changes.

[0181] During the overall driving risk assessment, if the state recovery deviation data does not indicate a decline in recovery capability, the normal state can be maintained; if the state recovery deviation data indicates a decline in recovery capability but the driving response margin has not yet reached the risk condition, the state can be identified as a state of concern. When the driving response margin reaches the risk condition and enters this step, it is identified as an immediate driving risk state, and the risk state is adjusted based on the state recovery changes obtained from encrypted monitoring and the updated driving response margin.

[0182] When the driving response margin is not higher than the risk margin boundary, it indicates that there is not enough remaining response space between the driver's current response capability and the time constraint of the target driving needs. When encrypted monitoring is started, it can be identified as an immediate driving risk state.

[0183] In an immediate driving risk state, if the state recovery changes obtained from encrypted monitoring show a recovery trend, and the updated driving response margin is again higher than the corresponding risk margin boundary, it indicates that the driver's state recovery ability is recovering to a normal level, and the response space for completing the current driving task has increased again. The driving risk assessment result can be adjusted from an immediate driving risk state to a state of concern. When the subsequent state recovery characteristics further recover to the range corresponding to the personal recovery baseline, and the driving response margin remains sufficient, it can be further restored to a normal state.

[0184] Conversely, when the state recovery changes obtained from encrypted monitoring show a continuous deviation from the state or a continuous downward trend, and the updated driving response margin is still not higher than the corresponding risk margin boundary, it indicates that the driver's state recovery ability has not been effectively recovered. At the same time, the current road driving task still has a strong time constraint on the driver's response ability. In this case, a continuous driving risk state is determined to be formed.

[0185] Therefore, the driving risk assessment results may include a normal state, a state of concern, an immediate driving risk state, and a continuous driving risk state. Specifically, the normal state indicates that the state recovery deviation data does not indicate a decline in recovery capability; the state of concern indicates that the state recovery deviation data indicates a decline in recovery capability, but the driving response margin has not yet reached a risk condition; the immediate driving risk state indicates that the driving response margin has reached a risk condition; and the continuous driving risk state indicates that the driving response margin is within the risk condition range, while intensive monitoring shows that the driver's state recovery capability has not been effectively restored.

[0186] For example, in one exemplary scenario, the driver's current driving response margin has decreased to 0.4 seconds, reaching the risk margin boundary corresponding to the deceleration of the vehicle in front. The system then initiates encrypted driving state monitoring. If the recovery time of subsequent consecutive state recovery events gradually decreases from 1.5 seconds to 1.1 seconds and then to 0.9 seconds, and the stability of the forward gaze gradually approaches the driver's normal recovery level, while the updated driving response margin is again above the risk margin boundary due to the driver having already decelerated or the distance to the vehicle in front increasing again, then the state recovery change is determined to represent a recovery trend, and the corresponding driving risk state is reduced.

[0187] Conversely, if during encrypted monitoring the driver's recovery time remains significantly higher than the individual baseline recovery time, and further shows an increase in recovery time and a decrease in forward visibility stability, while the driving response margin corresponding to the current or subsequent target driving needs remains below the risk margin boundary, then the change in recovery time is determined to be a continuous downward trend, and a continuous driving risk state is generated. The above values ​​and changes are illustrative examples and not fixed judgment parameters.

[0188] Through the above processing, this step does not directly conclude a persistent high-risk situation after detecting a single decline in driving condition. Instead, it first uses whether the driving response margin reaches the risk condition corresponding to the target driving requirements as the trigger for encrypted monitoring. Then, it uses continuous driving state data obtained with increased sampling frequency to determine changes in the driver's recovery ability. These changes in recovery ability, along with the driving response margin, are used together to form the driving risk assessment result. Thus, the driving risk result simultaneously reflects the time urgency of the current driving task and whether the driver's recovery ability is recovering or continuously declining, thereby completing a closed-loop process from driver state monitoring and driving task response capability matching to continuous confirmation of driving risk.

[0189] Example 2

[0190] This embodiment provides a ride-hailing driving risk assessment system based on driver status monitoring. The system is used to implement the ride-hailing driving risk assessment method in the above embodiment, and includes a data acquisition module, a status recovery module, a status deviation module, a driving demand module, a response analysis module, and a risk assessment module.

[0191] The data acquisition module is used to acquire driving status data, vehicle data, operational data, environmental data, and historical driving data. The driving status data can be acquired by a driver monitoring device; the vehicle data can be acquired by a vehicle control system or vehicle bus; the operational data can be acquired by a ride-hailing operation terminal and navigation terminal; the environmental data can be acquired by a vehicle environmental perception device and navigation map; and the historical driving data is stored in correspondence with the current driver. The data acquisition module can also correlate data from different sources according to time information for subsequent modules to access.

[0192] The state recovery module is used to determine the current operational task based on operational data and vehicle data, match driving state data with the operational task, identify task-related state changes, and determine state recovery characteristics based on driving state data after the task-related state changes have ended. Specifically, the state recovery module can identify task-related state changes based on the observation area and effective time period corresponding to the current operational task, and determine the state recovery completion time through changes in the driver's forward gaze and head posture, thereby obtaining the state recovery time and the degree of forward gaze stability.

[0193] The state deviation module is used to establish a personal recovery baseline for the current driver based on historical driving data, and compares the state recovery characteristics with the personal recovery baseline to determine the state recovery deviation data. The personal recovery baseline can be categorized and stored according to operational tasks, vehicle operating status, and road conditions. After the current state recovery characteristics are generated, the state deviation module determines the corresponding target recovery baseline from the personal recovery baseline, and judges whether the driver's state recovery ability has continuously deviated based on the difference between the state recovery time and the stability of forward vision.

[0194] The driving demand module is used to determine the target driving demand and the corresponding available response time based on environmental and vehicle data when the state recovery deviation data indicates a decline in the driver's recovery ability. Specifically, candidate driving demands within a short time range can be determined based on the target ahead of the vehicle, the vehicle's position, vehicle speed, relative motion relationship, and road structure, and driving tasks with stronger constraints on the driver's response time are prioritized as the target driving demand.

[0195] The response analysis module is used to determine the driving response time based on the target driving demand, state recovery deviation data, and historical driving data, and to determine the driving response margin based on the driving response time and available response time. Specifically, the current attention recovery time can be determined based on the state recovery deviation data, and the corresponding manipulation response time can be determined based on the target driving demand and historical driving data. The driving response time is then determined by comparing the attention recovery time and the manipulation response time; finally, the driving response time is compared with the available response time to determine the driving response margin corresponding to the current target driving demand.

[0196] The risk assessment module is used to perform encrypted monitoring of driving status when the driving response margin reaches a risk condition, and to generate a driving risk assessment result based on the driving response margin and the changes in state recovery obtained during the encrypted monitoring process. Specifically, a corresponding risk margin boundary can be determined according to the target driving needs. When the driving response margin is not higher than the risk margin boundary, the sampling frequency of driving status data is increased, and it is continuously judged whether the driver's state recovery ability is recovering to the individual's normal level or continuing to decline, thereby forming a corresponding driving risk assessment result.

[0197] The aforementioned modules can be implemented by the onboard computing unit or the processor in the ride-hailing operation safety platform, or they can be implemented collaboratively by the onboard unit and the server. The modules communicate with each other via data interfaces, transmitting driving status data, status recovery characteristics, status recovery deviation data, target driving requirements, driving response margin, and driving risk assessment results, thereby completing the continuous assessment of ride-hailing driving risks.

[0198] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit the scope of protection of this application. Those skilled in the art can modify the above embodiments or make equivalent substitutions for some or all of the technical features without departing from the technical concept of this application. The resulting technical solutions should still fall within the scope of the technical solutions disclosed in this application, and the scope of protection of this application shall be determined by the scope defined in the claims.

[0199] The technical solution of this application can be implemented in hardware, software, or a combination of both. When implemented in software, it can form a computer program product, which can be stored in a computer-readable storage medium, including but not limited to solid-state drives, flash memory, mobile storage devices, optical discs, ROM / RAM, and server storage media. The computer program product includes computer-readable instructions, which, when executed by one or more processors, implement the methods described in the above embodiments.

[0200] The terms “first,” “second,” “third,” and “fourth” used in this application are only used to distinguish different objects and do not indicate a specific order or degree of importance; the terms “including” and “comprising” are used to indicate the presence of the stated features, steps, operations, elements, or components and do not exclude the presence or addition of other features, steps, operations, elements, or components.

[0201] Unless the context explicitly defines it otherwise, the singular forms “a,” “an,” “the,” etc., used in this application include the plural forms; “and / or” means any combination of one or more items listed in connection with the application. The terminology used in this specification is for describing particular embodiments only and is not intended to limit the scope of protection of this application.

Claims

1. A method for assessing ride-hailing driving risks based on driver status monitoring, characterized in that, include: Acquire driving status data, vehicle data, operational data, environmental data, and historical driving data; The operation task is determined based on the operation data and vehicle data, the driving status data is matched with the operation task to obtain task-related status changes, and the status recovery characteristics are determined based on the driving status data after the task-related status changes end. A personal recovery baseline is established based on the historical driving data, and the state recovery characteristics are compared with the personal recovery baseline to obtain state recovery deviation data. When the state recovery deviation data indicates a decline in recovery capability, the target driving demand and available response time are determined based on the environmental data and vehicle data. The driving response time is determined based on the target driving requirements, state recovery deviation data, and historical driving data, and is compared with the available response time to obtain the driving response margin. When the driving response margin reaches the risk condition, encrypted monitoring of the driving state is performed to obtain the state recovery change. The driving risk assessment result is generated by combining the driving response margin and the state recovery change.

2. The method according to claim 1, characterized in that, Based on the driving state data after the completion of the task-related state changes, state recovery characteristics are determined, including: Extract forward gaze state and head posture change data from the driving state data after the completion of the task-related state changes to obtain recovery monitoring data; Based on the recovery monitoring data, the forward gaze state and head posture changes are continuously judged to determine the time when the state recovery is completed. The state recovery time is determined based on the end time of the task-related state change and the completion time of state recovery, and the state recovery characteristics are determined in conjunction with the recovery monitoring data.

3. The method according to claim 2, characterized in that, Based on the restored monitoring data, continuous determination is made of forward gaze status and head posture changes, including: The first recovery moment when the driver's line of sight re-enters the forward road area is determined based on the recovery monitoring data; Starting from the first recovery time, the forward gaze state and head posture changes are continuously monitored to determine the continuous recovery interval that meets the stable recovery conditions. The time when state recovery is completed is determined based on the continuous recovery interval.

4. The method according to claim 3, characterized in that, Determine the continuous recovery interval that meets the stable recovery criteria, including: The forward gaze continuity status and head posture deviation status are determined based on the recovery monitoring data. When the forward gaze is continuously maintained and the forward stability condition is met, and the head posture deviation condition is met, the corresponding time period is determined as a candidate recovery time period. Consecutive candidate recovery periods are merged to determine the continuous recovery interval that meets the stable recovery conditions.

5. The method according to claim 1, characterized in that, The recovery characteristics are compared with the individual recovery baseline to obtain recovery deviation data, including: Based on the operational tasks, vehicle data, and environmental data, a target recovery baseline is determined from the individual recovery baseline; The recovery characteristics are compared with the target recovery baseline to obtain recovery difference data; Based on the recovery difference data, the deviation of the current state recovery from the target recovery baseline is determined, and the state recovery deviation data is obtained.

6. The method according to claim 5, characterized in that, Determining the deviation of the current recovery state from the target recovery baseline based on the recovery difference data includes: Based on the recovery difference data, determine the differences in state recovery time and forward line of sight stability corresponding to continuous state recovery events; The continuous deviation of consecutive state recovery events from the target recovery baseline is determined based on the state recovery time difference; When the state recovery time of a continuous state recovery event is consistently higher than the baseline state recovery time corresponding to the target recovery baseline, and the forward line-of-sight stability is consistently lower than the baseline forward line-of-sight stability corresponding to the target recovery baseline, the state recovery deviation data is determined.

7. The method according to claim 1, characterized in that, Determine the driving response time and compare it with the available response time to obtain the driving response margin, including: The time difference between the two is determined based on the available response time and the driving response time; The driving response margin corresponding to the target driving requirement is determined based on the time difference.

8. The method according to claim 7, characterized in that, Determining the driving response margin corresponding to the target driving demand based on the time difference includes: Determine the corresponding response margin boundary based on the target driving requirements; The time difference is compared with the response margin boundary to determine the positional relationship of the time difference relative to the response margin boundary; The driving response margin corresponding to the target driving requirement is determined based on the positional relationship and time difference.

9. The method according to claim 1, characterized in that, When the driving response margin reaches a risk condition, encrypted monitoring of the driving status is performed, including: Determine the corresponding risk margin boundary based on the target driving requirements; The driving response margin is compared with the risk margin boundary. When the driving response margin is not higher than the risk margin boundary, it is determined that the driving response margin has reached the risk condition. After the aforementioned risk conditions are met, the sampling frequency of driving status data is increased, and the state recovery change is determined based on the encrypted driving status data.

10. A ride-hailing driving risk assessment system based on driver status monitoring, characterized in that, include: The data acquisition module is used to acquire driving status data, vehicle data, operational data, environmental data, and historical driving data. The status recovery module is used to determine operational tasks and task-related status changes, and to determine status recovery characteristics; The state deviation module is used to establish an individual recovery baseline and determine state recovery deviation data based on the state recovery characteristics. The driving demand module is used to determine the target driving demand and available response time based on the environmental data and vehicle data. The response analysis module is used to determine the driving response time and obtain the driving response margin based on the driving response time and available response time. The risk assessment module is used to perform encrypted monitoring of driving status and generate driving risk assessment results when the driving response margin reaches the risk condition.

Images